通过self .__ setattr__进行迭代生成的实例属性。 有没有更好的办法？

Question

When you setting attributes to an instance from some iterable like this:

class DogWith100Legs():
    def __init__(self, legs_colors):
        for leg_i, color in legs_colors.items():
            self.__setattr__(leg_i, color)

legs_colors = {'leg_1': 'white', 'leg_2': 'brown', ... , 'leg_100': 
'black'}
doge = DogWith100Legs(legs_colors)
print(doge.leg_100)

it works fine. But PyCharm will highlight leg_100 with a warning: "Unresolved reference 'leg_100' for class 'DogWith100Legs'.

Why i want to do this? I want to create tree-like object. Like BeautifulSoup do. But bs4 does not have such a problem in PyCharm

Also, there are comments in similar questions (like PyCharm warns about unresolved attribute for dict generated attributes ), where people say, that it is a bad practice

What will be a good practice? Or may be you know a good way to calm down Pycharm? (Not disabling Pycharm check)

Answer 1

PyCharm (and every other IDE) will give you warnings/errors because in __init__ you have to declare every member of the object.

Because you are passing a dict, probably the best thing to do is to instantiate a dict into your class too, and then just get the value you need with the key self.my_dict['leg_x'] and not self.leg_x .

class DogWith100Legs():
def __init__(self, legs_colors):
    self.legs_colors = legs_colors.copy()

legs_colors = {'leg_1': 'white', 'leg_2': 'brown', 'leg_100': 'black'}
doge = DogWith100Legs(legs_colors)
print(doge.legs_colors['leg_100'])

Answer 2

It depends on intended usage.

1 If you are going to access the data directly by name, and there is no obvious relationship between items.

print(dog.leg_32)
print(dog.leg_5)

Only in this case is __setattr__ fine.

2 You plan to retrieve the data indirectly, and/or there is some relationship (eg previous, next):

i = 25
print(dog.leg[i])
print(dog.leg[i+1])

In this case all data should be placed into a container ( list , dict are the most common ones). Most probably this is what you want.

Answer 3

I would rethink what you are trying to do a little bit - maybe even check out the BeautifulSoup source code to see how they do it.

In general though, there is no reason (that I can think of) to dynamically define instance attributes like this. You should simply have a leg dictionary or list within your Dog objects, like so:

def __init__(self, legs_colors: dict):
    self.legs_colors = legs_colors.copy()

Instance attributes should be individually important, but I doubt that you plan to use leg_1 differently than you plan to use leg_72 , especially since you can't guarantee that either of these attributes will exist with the way you have defined __init__ .

Also, when you later access these leg attributes, you will likely want to do so dynamically. In which case you would have to resort to something like:

for leg in [f'leg_{i}' for i in range(100)]
    color = getattr(doge, leg)
    # do something with leg and color

instead of the simpler, clearer, and less smelly:

for leg, color in doge.legs_colors.items():
    # do something with leg and color

You can still program different interfaces for accessing your legs_colors dictionary, if you'd like to. For example, you could overwrite __getitem__ on your Dog class if you want to be able to do doge[i] directly.